The probability integral method is the most commonly used mining subsidence prediction model, but it is only applicable to ordinary geological mining conditions. When the loose layer in the geological mining conditions where the mining face is located is too thick, many inaccurate phenomena will occur when the movement deformation value is predicted by the probability integral method. The most obvious one is the problem that the predicted value converges too fast compared with the measured value in the edge of the sinking basin. In 2012, Wang and Deng proposed a modified model of probability integral method for the marginal errors in the model of probability integral method and verified the feasibility of the method through examples. In this paper, the method is applied to the prediction of surface movement under thick and loose layers after modified. Through practical application, it is found that due to the angle between the working face and the horizontal direction, the average mining depth in the strike direction is different from the average mining depth in the inclined direction, and the main influence radius of the two main sections are often. Therefore, based on this problem, this paper divides the main influence radius into trend and tendency and adjusts the parameters in the model to find the rules of the parameters. The original method uses a dynamic scale factor to adjust the predicted shape of the graph by adjusting the sinking coefficient. This study is aimed to set the scale factor to 0.5 and fix the value of the sinking factor, and propose to adjust the integral range and then adjust the shape of the graph to make it more in line with the actual measurement situation.

The probability integral method (PIM) is the main method for mining subsidence prediction in China. Parameter errors and model errors are the main sources of error in the application of the probability integral method. There are many surface subsidence problems caused by coal mining. In order to improve the accuracy and operating efficiency of the genetic algorithm (GA) in calculating the parameters of the PIM, this paper proposes an improved genetic algorithm (IGA) by adding the dynamic crossover and mutation rate to the traditional GA. Made improvements to the shortcomings of random crossover and mutation rate of all individuals in the population in the original algorithm.Through simulation experiments, it is confirmed that the IGA improves the calculation efficiency and accuracy of the traditional GA under the same conditions.The IGA has higher accuracy, reliability, resistance to gross interference and resistance to missing observation points. This method is obviously superior to direct inversion and conventional optimization inversion algorithms, and effectively avoids the dependence on the initial value of the probabilistic integral method parameter.